Political conflict in Bismarck's Germany - CORE

Political conflict in Bismarck’s Germany: An analysis of

parliamentary voting, 1867-1890

Frank M. Häge1

Department of Politics and Public Administration

University of Limerick

Abstract

Imperial Germany is a prominent historical case in the study of Western Europe’s political

development. This paper investigates the number and content of political conflict dimensions

from the foundation of the modern German state in 1866 to the end of Bismarck’s reign as

Chancellor in 1890. Methodologically, it applies dimension-reducing statistical methods to a

novel dataset of content-coded parliamentary roll call votes. The analysis suggests that the

emergence of the Catholic Centre Party in 1871 permanently transformed the conflict space

from a single liberal-conservative divide to a two-dimensional space that distinguished

positions on socio-economic issues and regime matters, respectively. The fact that positions

on redistributive and regime issues were not aligned implies that theories stressing economic

inequality as a driver for regime change are of limited applicability. Instead, the case of

Imperial Germany highlights the importance of cross-cutting non-economic societal

cleavages and the role of societal and political organisations in drawing attention to and

perpetuating these divisions.

Keywords

Dimensionality, political space, roll call votes, Reichstag, Imperial Germany

1 Email: [email protected]; web: www.frankhaege.eu.

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by University of Limerick Institutional Repository

1

Political conflict dimensions in Bismarck’s Germany

Germany’s political development before World War I forms the backdrop for a number of

influential theories of democratization (Moore 1966), party development (Michels 1911), as

well as party system creation and change (Lipset and Rokkan 1967). Representing the

paradigmatic case of the ‘late democratizer’ in Western Europe, even contemporary scholars

of democratization feel compelled to demonstrate that their theories can provide an

explanation for the longevity of the Kaiser’s regime (e.g., Acemoglu and Robinson 2006: 67,

200–201; Ansell and Samuels 2014: 28–29, 52–54). This study contributes to our knowledge

about German political development by identifying the dimensionality and content of the

political conflict space during the founding years of the modern German state. It does so by

analysing a novel data set of roll call votes taken in the Reichstag of the North German

Confederation (1867-1871) and during the first seven legislative terms of the Reichstag of the

German Empire (1871-1890). The study period covers Bismarck’s entire term as Chancellor,

which is often seen as a distinct era in the history of Imperial Germany (e.g., Biefang 2012:

17–18). The study complements the early dimensional analysis by Smith and Turner (1981)

of parliamentary voting in Wilhelmine Germany as well as more recent studies by Debus and

Hansen (2010; Hansen and Debus 2012) on parliamentary voting in the Weimar Republic.

However, the study does not only increase our descriptive knowledge about the nature

of political competition in an important historical case, it also contributes to the

methodological debate about the interpretation of roll call analyses and the theoretical

discussion about the causes of regime stability and change. Methodologically, the analysis

illustrates the importance of taking the policy agenda into account when interpreting the

results of a statistical scaling analysis. Apparent changes in the number of dimensions can be

due to the genuine appearance or disappearance of particular conflict constellations as a result

of the entrance or exit of actors or issues, or artificially induced by the legislative agenda

restraining votes to policy issues related to only a subset of the dimensions. In practice, only

substantive knowledge about the content of votes and the historical context of the case can

help distinguish between these two scenarios. This is a particularly important issue when

analysing parliamentary voting in historical cases with comparatively low levels of legislative

productivity and a small number of associated roll call votes. In such circumstances, the

results of scaling analyses are particularly sensitive to agenda effects.

Theoretically, the study results inform the debate on economic inequality as a cause of

democratization (e.g., Acemoglu and Robinson 2006; Ansell and Samuels 2014; Boix 2003).

2

A shared assumption of these political economy approaches to democratisation is that regime

preferences are induced by economic interests. If that was the case, political conflict

dimensions should be aligned with and reducible to divergent socio-economic cleavages.

However, the findings of the scaling analysis show that such an alignment did not generally

exist in 19th

century Germany. For most of the time period studied, positions of party groups

varied along two separate dimensions, one relating to socio-economic issues and one to

regime matters. This separation was mainly due to the emergence of the Centre Party in 1871,

which can ultimately be traced back to a religious cleavage between the Catholic minority

and the Protestant majority in Germany. Given the significant influence of this non-economic

factor in structuring political competition, theories based on economic inequality as primary

causal variable for democratization cannot provide convincing explanations for the longevity

of this authoritarian regime.1

The political system of Imperial Germany

The North German Confederation was established under Prussian leadership in 1867, after

Prussia had defeated Austria in the ‘German War’ of 1866. A few years later, another

victorious war of Prussia, this time against France, paved the way for the accession of the

South German states to the Confederation and for the foundation of the German Empire. The

constitution of the German Empire of 1871 was largely inherited from the North German

Confederation. One of the constitution’s most progressive features was the provision of equal

suffrage for all male citizens of 25 years’ age or older. However, as an intentional

‘corrective’, the constitution banned the payment of allowances to deputies in an effort to

restrict the passive suffrage (Butzer 1999: 29). It also established a strict separation-of-power

system, consciously designed to prevent a parliamentarization of the regime (Rauh 1973: 48).

The Kaiser remained in full control of the military, and the Reichstag’s yearly budget

approval right did not extend to the military budget. The Reichstag shared legislative powers

with the Bundesrat, which represented the governments of the kingdoms and principalities in

this federal monarchy. Contrary to a common misperception, the Reichstag was able to

initiate legislation (Anderson 2000: 10); and in practice, the approval of bills involved

reaching compromises between the Reichstag and the Bundesrat, similar to bicameral

bargaining in current-day legislatures.2

1 For a similar conclusion regarding the value of theories of economic inequality for explaining Argentina’s

democratic breakdown in the 1920s, based on a similar methodological approach, see Alemán and Saiegh’s

(2014). 2 For case study examples, see the descriptions of legislative decision-making in Rauh (1973) or Butzer (1999).

3

By occupying several important posts between 1867 and 1890, Bismarck held a

particularly elevated position in this system. As Chancellor, Bismarck chaired the Bundesrat

and was the only politically responsible member of the imperial government (Reichsleitung).

Government departments were not led by ministers but state secretaries, who were

subordinate to the Chancellor. At the same time, Bismarck was the prime minister and

foreign secretary of Prussia. In this latter function, he commanded Prussia’s votes in the

Bundesrat. Prussia’s votes were sufficient to veto any constitutional change and, together

with the votes of various micro-state enclaves that were completely dependent on Prussia

(Rauh 1973: 60–61), could determine legislative decisions in this body. Although Bismarck

enjoyed strong public support, his power and influence was ultimately dependent on the trust

and backing by the Kaiser and King of Prussia, who could unilaterally appoint and dismiss

the Chancellor or members of the Prussian government.

At the national level, the parties that competed for seats in the Reichstag can be divided

into four broad camps. The liberal camp consisted of the left-leaning German Progress Party

(Deutsche Fortschrittspartei), renamed to German Liberal Party (Deutsch-Freisinnige Partei)

in 1884, and the right-leaning National Liberal Party (Nationalliberale Partei). In contrast to

the Left Liberals, the National Liberals were willing to compromise their political liberal

principles as long as progress was made on establishing a liberal economic order in a unified

nation state. During the first ten years of Bismarck’s reign as Chancellor, the National

Liberals were his primary partner in the Reichstag. The conservative camp consisted of the

Free Conservative Party (Freikonservative Partei) and the German Conservative Party

(Deutsch-Konservative Partei).3 The Free Conservatives, mostly high aristocracy and senior

bureaucrats, were ardent supporters of Bismarck’s policies throughout his term of office

(Stalmann 2000). The German Conservatives, which included more members of the landed

Prussian aristocracy, the infamous Junkers, were more critical of Bismarck’s early policies

regarding national unification and the implementation of largely liberal economic ideas. The

Catholic camp consisted of the Centre Party (Zentrum), often supported by representatives of

regional and national minority groups (i.e. Poles, French, and Guelphs). Some of its leaders

had already been members of the diverse Particularist party group in the Reichstag of the

North German Confederation, but the Centre Party was only formed in 1870. It entered the

Reichstag after the foundation of the German Empire in 1871. Finally, the Socialists

3 The latter adopted the adjective ‘German’ only in 1876, but to avoid confusion with the Free Conservatives

and references to the broader conservative camp encompassing both party groups, the term is also used to refer

to the Conservative Party pre-1876.

4

constitute the last camp. Before their merger in 1875 under the label of Socialist Worker’s

Party (Sozialistische Arbeiterpartei), this camp was composed of two small factions, which

sometimes fiercely opposed each other, the General German Workers’ Association

(Allgemeiner Deutscher Arbeiterverein) and the Social Democratic Worker’s Party

(Sozialdemokratische Arbeiterpartei).

Figure 1 shows the share of Reichstag seats of these party groups over time. Party

groups are roughly ordered along their position on a general liberal-conservative dimension.

Conservative party groups are at the top of the figure, and liberal ones at the bottom. Cells are

shaded according to the share of seats a party group held during a legislative term. Although

the table indicates considerable fractionalization, it also shows that five groups dominated the

Reichstag for most of the period: the two conservative and the two liberal parties, and since

1871, the Centre Party. The table also shows that the Socialists were a fringe group that

played no decisive role in the Reichstag during that period of time.4

At this point, it should be noted that the fragmented multiparty nature of legislative

competition in Germany’s national parliament differed from electoral competition at the

district level. Partly due to Germany’s electoral system with run-off elections in single-

member districts, party political competition at the local level was often limited to two or

three parties; and the identity of these parties often depended on local socio-economic, ethnic,

or religious cleavages. In the most extreme cases, party competition was not only more

limited but almost completely absent, with the election of the candidate of a particular party

being virtually a certainty. The Centre Party was particularly successful in that respect, but a

sizeable number of Polish and German Conservative seats were similarly secure (Schauff

1973: 307). Thus, the analysis presented here describes party political competition at the

national level, resulting from the electoral aggregation of quite disparate political conflict

constellations in electoral districts at the local level.

Figure 1 allows us to make some inferences about the potential influence of different

groups and possible majority constellations at the national level in the Reichstag. First, the

government could never rely on an exclusively conservative majority. The two conservative

party groups reached their highest combined seat share of 34.6% in the Reichstag of the

North German Confederation of 1867. Second, the North German Confederation was

essentially a bipolar system, separating the conservative from the liberal camp, but neither

camp commanded a clear majority on its own. Third, the newly formed Centre Party entered

4 A partial exception is the sixth term of the Reichstag, in which the Socialists held 6 per cent of the mandates.

5

the first Reichstag of the German Empire immediately as the second largest party and from

1874 continuously won about a quarter of the seats. Considering that the minority groups

usually supported the Centre Party, the effective share of Reichstag votes commanded by this

party group was probably closer to one-third. Fourth, the National Liberals were the

dominant party group until 1878. They reached the peak of their parliamentary representation

of 38% during the second Reichstag term, after the Prussian government had thrown its

support behind National Liberal rather than German Conservative candidates during the

election of 1874.

Figure 1 Share of Reichstag seats of party groups by legislative term

Notes: Minorities include French, Polish, and Guelphs. Besides the German Progress Party/German Liberal

Party (Deutsche Fortschrittspartei/Deutsch-Freisinnige Partei), Left Liberals include the Free Association (Freie

Vereinigung, 1867-1871), the German People’s Party (Deutsche Volkspartei, 1871-1890), and the Liberal

Association (Liberale Vereinigung, 1881-1884). Before 1877, Socialists include the General German Workers’

Association (Allgemeiner Deutscher Arbeiterverein) and the Social Democratic Workers’ Party

(Sozialdemokratische Arbeiterpartei). These two groups merged in 1875 to form the Socialist Workers’ Party

(Sozialistische Arbeiterpartei). NDRT = Reichstag of the North German Confederation; RT = Reichstag of the

German Empire. See Table A1 in the online appendix for a more detailed numerical breakdown.

6

However, in 1878, Bismarck dissolved the Reichstag prematurely, blaming the liberal party

groups for the rejection of his first anti-socialist law bill. As a consequence, public support

for liberals of all strands dropped. Furthermore, during the fourth term, the left wing of the

National Liberals seceded from the party group over disagreements regarding the support of

Bismarck’s protectionist and repressive policies. The fifth and sixth Reichstag had the most

left-leaning and oppositional composition. The National Liberals only made significant gains

again in the elections of 1887, after another premature dissolution of the Reichstag by the

government. This time, Bismarck had blamed the Centre Party, the Left Liberals, and

Socialists for rejecting his proposal for another seven-year military budget, claiming that

increases in military spending were necessary to avert an allegedly imminent attack by

France. Forming the so-called ‘Kartell’ electoral coalition with Free and German

Conservatives, National Liberal election candidates enjoyed the support of a united right as

well as the Prussian bureaucracy. However, even this government-friendly alliance broke

apart in 1890 because of its members’ refusal to compromise over the renewed extension of

the anti-socialist law.

Based on the historical record, it is difficult to develop firm expectations about the

dimensionality of the party system’s conflict space or the content of its dimensions,

especially as both might also have changed over time. With respect to its dimensionality,

contemporaries and historians referred to the Centre Party and the National Liberals often as

the ‘middle parties’ (e.g., Nipperdey 2013: 513), implying that they occupied ground in-

between ideological poles. At the same time, they also suggest that Bismarck was able to

govern with varying majorities (e.g., Pflanze 1982: 572), which indicates the existence of a

multi-dimensional conflict space. In general, given the high fragmentation of the party system

and the variety of underlying cleavages, it seems likely that the conflict space was

multidimensional, but expectations about the exact number of dimensions and their content

are hard to discern a priori.

Collection and coding of Reichstag roll call votes

To identify the number and the content of political conflict dimensions in Bismarck’s Reich,

the study conducts a scaling analysis of roll call votes taken in the Reichstag between 1867

and 1890. In the Reichstag, the standard way of voting was to stand up or remain seated. A

proposal for a vote by roll call required the support of at least 50 deputies, which constituted

7

quite a large threshold.5 Before the introduction of voting by division through an amendment

of the rules of procedure in 1874, votes that were too close to call also immediately triggered

a roll call.6 In practice, these thresholds ensured that recorded votes were called on important

and divisive topics that engaged a substantial number of legislators. Possible selection biases

are a fundamental concern for studies of roll call voting (e.g. Carrubba et al. 2006). However,

to the extent that the probability of calling a roll call vote is positively associated with

political conflict, the selective nature of roll call votes actually results in a more informative

sample for the purposes of identifying conflict lines in the legislature; and given the

distribution of mandates across party groups, none of the three main camps (conservatives,

liberals, Catholics) in the Reichstag were prevented from requesting a roll call vote during

any of the legislative terms.

The source for the voting information is the overview of roll call votes in Appendix A

of the ‘General Register for the Stenographical Reports of all Reichstag Sessions from 1867

to 1895’ (Reichstagsbureau 1896). After downloading the digitized version of the General

Register from the website of the Bavarian State Library, optical character recognition (OCR)

software was used to convert the PDF images of the roll call vote appendix into machine-

readable text. To identify the full population of legislators and disambiguate their party group

affiliation and electoral district at the time of a vote, the vote data was linked to biographical

information from the online database Biorab-Kaiserreich, which is hosted by the Centre for

Historical Social Research at the GESIS Leibnitz Institute for the Social Sciences. Before

combining the two datasets through a custom-made record linkage algorithm, information

from both sources was extracted through computer scripts developed in Python. Several steps

in the data collection process required extensive manual review and corrections based on the

consultation of historical primary and secondary sources.7

To assess possible agenda changes in the roll call data over time, the policy area of each

roll call was coded from its brief description in the General Register. The description is

sufficient to identify the general policy area of a proposal, but not for identifying the specific

rationale behind amendments to articles or paragraphs within a proposal. To reliably code

policy content at the level of individual amendments, detailed case knowledge about the

context of hundreds of individual votes would have been required. Absent such knowledge,

5 The total number of Reichstag members was 297 in the North German Confederation (1867-1871), 382 in the

first term of the Reichstag of the German Empire (1871-1874), and 397 in the remaining legislative terms, after

15 new members had joined in 1874 to represent districts of newly annexed Alsace-Lorraine. 6 Stenographische Berichte des Reichstags 1874 II/1, vol. 2 [27], 680ff., 9 April 1874.

7 Further source information, the details of the data extraction and coding process, as well as a description and

discussion of the distribution of votes across types of decisions are given in the online appendix.

8

the coding takes the stated policy goals of the proposal as a whole at face value. However, it

needs to be acknowledged that this coding procedure is likely to understate somewhat the

number of votes that had implications for civil rights and liberties, the relative power of the

Reichstag vis-à-vis other institutions, or the relative power of the Reich vis-à-vis its

component states. In terms of policy categories, the topic coding scheme of the Comparative

Agendas Project was used.8 Despite being widely used in the study of comparative policy-

making, this seems to be the first time that the scheme is applied to a historical case.

In general, the roll call votes in the dataset are mostly about the operation and

organisation of the state (19 per cent), foreign trade regulation (16 per cent), military

organisation and defence spending (13 per cent), law and crime (12 per cent), the regulation

of domestic commerce (11 per cent), civil rights and liberties (10 per cent), macroeconomic

policy (6 per cent), and social welfare (5 per cent). The fact that roll call votes are

concentrated in 8 out of 20 possible policy categories shows how limited the agenda diversity

of the Reichstag was compared to current-day legislatures. The Reichstag was in session only

about 3 to 6 months a year; and the involvement of the state in social and economic matters

only started to build up in the 1880s. Thus, both the capacity and demand for legislative

productivity was comparatively low.

Figure 2 shows that the policy area concentration of roll call votes is even more

pronounced within individual legislative terms. This concentration of the agenda is not a

methodological artefact but reflects a real focus of the Reichstag’s limited attention on a

single or a small number of particularly important and comprehensive pieces of legislation.

These prominent pieces of legislation often involved a large number of amendments decided

through roll call votes. As long as there is a sizable share of roll call votes in policy areas

relating to other conflict dimensions, a strong concentration of attention on a single policy

area is not problematic for identifying the dimensionality of the conflict space or the content

of conflict dimensions. For the purposes of this study, selection effects of the agenda are only

consequential when they focus the attention of the Reichstag on policy areas that relate

exclusively to a subset of the dimensions structuring the conflict space. For example, the

analysis below shows that the scaling results for the second legislative term of the Reichstag

suffer from the effects of a selective agenda; not because its agenda was focused on any

8 For the project, see http://www.comparativeagendas.info; for a list of all topic codes, see

https://dl.dropboxusercontent.com/u/552280/CAPMasterCodebook/MasterCodebookTopics.csv (both last

accessed 11 April 2016).

9

single policy area, but because its agenda was focused on a set of policy areas that all relate

to a single dimension of a multidimensional conflict space.

From a historical perspective, it is interesting that Figure 2 shows a clear pattern in

attention to different policy areas over time. During the so-called ‘liberal era’ of the North

German Confederation (NDRT) and the first three legislative terms of the Reichstag (RT 1 to

RT 3), roll call votes focused to a large extent on measures aimed at clarifying the role and

operation of government institutions, as well as the establishment of a functioning federal

administration. Attention was also devoted to developing a penal code (Law, Crime, and

Family Issues in NDRT and RT 2) and a harmonised court system (RT 2). In contrast, after

what some historians have called the ‘second founding’ of the Reich in 1878 (Barkin 1987;

Gerschenkron 1943), the focus switched from state-building measures to economic and social

policies. In particular, trade tariffs (Foreign Trade in RT 4, 6, and 7), domestic taxes

(Macroeconomic Issues in RT 4, 5, 6, 7), and a new system of social welfare insurance

(Social Welfare in RT 4, 5, and 7) were high on the agenda. Finally, three policy areas were

the subject of roll call votes throughout the period studied here: civil rights and liberties

(especially repressive measures against the Catholic Church during the Culture War and

against Social Democrats after the anti-socialist law was first introduced in 1878), the

organisation and financing of the military, and the regulation of domestic commerce. The fact

that much of the regulation of domestic commerce in the 1880s (RT 5 and 7) was about

‘correcting’ the liberal policies adopted in the North German Confederation in the late 1860s

(NDRT) is another telling sign of the programmatic reorientation of German economic policy

after 1878.

10

Figure 2 Share of roll call votes in different policy areas by legislative term

Notes: Policies were classified according to the classification scheme of the Comparative Agendas Project

(http://www.comparativeagendas.info). NDRT = Reichstag of the North German Confederation; RT = Reichstag

of the German Empire. See Table A2 in the online appendix for a full numerical breakdown.

In total, the data consists of 129,579 potential vote choices. These vote choices relate to 322

roll call votes in seven legislative terms by a total of 2884 members of the Reichstag.

Because the third Reichstag was dissolved prematurely, only nine roll call votes were taken

during that term. This low number of votes precludes a separate scaling analysis. Hence, this

legislative term had to be omitted from the sample. Table 1 shows that, in most other terms,

the number of roll call votes ranged between 30 and 50. Only the Reichstag of the North

German Confederation (NDRT) stands out for its exceptionally large number of roll call

votes (i.e., 81). The large number of roll calls in the NDRT is at least partly a result of the

high legislative productivity of this parliament (Pollmann 1985: 433, 451). It is also

noteworthy that the number of legislators per term is usually substantially higher than the

number of constituent members at the time, indicating substantial turnover of members

within a term.

11

Table 1 Roll call vote characteristics by legislative term

NDRT RT 1 RT 2 RT 4 RT 5 RT 6 RT 7

Yeah or nay vote 15710 12286 12383 11343 13020 9454 8114

0.57 0.62 0.70 0.62 0.65 0.69 0.63

Treated as missing 11668 7442 5299 6861 7002 4241 4756

0.43 0.38 0.30 0.38 0.35 0.31 0.37

Of which:

Abstained 20 37 49 30 52 118 352

0.00 0.00 0.00 0.00 0.00 0.01 0.03

Absent 11578 7360 5238 6797 6924 4121 4400

0.42 0.37 0.30 0.37 0.35 0.30 0.34

Not a member 70 45 12 34 26 2 4

at time of vote 0.00 0.00 0.00 0.00 0.00 0.00 0.00

Total 27378 19728 17682 18204 20022 13695 12870

(Proportion) 1.00 1.00 1.00 1.00 1.00 1.00 1.00

Number of roll calls 81 48 42 41 47 33 30

Number of legislators 338 411 421 444 426 415 429

Average turnout 0.58 0.63 0.70 0.63 0.65 0.70 0.66

(Standard deviation) 0.06 0.09 0.07 0.10 0.09 0.10 0.12

Average vote margin 0.19 0.22 0.26 0.22 0.2 0.21 0.29

(Standard deviation) 0.23 0.18 0.20 0.16 0.19 0.22 0.23

Notes: NDRT=Reichstag of North German Confederation, RT=Reichstag of German Empire

Table 1 also provides further information about the voting data that actually entered the

scaling analysis. In total, the scaling analysis is based on 82,310 substantive individual vote

choices (i.e. yeah or nay votes). As is standard in the literature, abstentions, absent members,

or instances where a legislator was not a member (yet or anymore) at the time of the roll call

are treated as missing values. In general, abstentions were extremely rare. They increased

somewhat towards the end of the study period, but still constituted only 1 and 4 per cent of

the substantive vote choices in the sixth and seventh legislative term of the Reichstag (RT),

respectively. Absenteeism was generally rather high. Average turnout ranged between 58 per

cent in the NDRT and 70 per cent in RT 2 and RT 6. Given the lack of parliamentary

allowances and the long distances to be travelled by legislators, low attendance rates were a

permanent problem in the Reichstag (Butzer 1999: 147). Although the number of roll calls in

each term is relatively small compared to many contemporary legislatures, they are generally

very informative. The average margin between yeah and nay votes ranged between 19

percentage points in the NDRT and 29 percentage points in RT 7. Thus, almost all roll calls

meet the minimum vote margin requirement for the scaling analysis.

12

The number and content of political conflict dimensions

To estimate legislators’ policy positions from the observed vote choices, I employ Poole’s

non-parametric Optimal Classification (OC) algorithm (Poole 2000; Poole 2005). Similar to

parametric alternatives, like Nominate (Poole and Rosenthal 2007) or Bayesian Item

Response Theory models (Clinton et al. 2004), OC assumes that the policy space is Euclidean

and that legislators vote sincerely based on single-peaked and symmetric utility functions. In

contrast to other methods, OC does not require specific assumptions about the functional

form of legislators’ utility functions or their distribution of voting errors (Armstrong et al.

2014: 249). In fact, when minimizing the total number of errors in its classification of vote

choices, OC treats all voting errors equally. In situations where the error rate is low or the

distribution of errors is unknown, OC has thus clear advantages (Armstrong et al. 2014: 265–

266).

At least two characteristics of voting in the Reichstag make the application of OC

preferable to parametric methods. Firstly, voting cohesion varies considerably across party

groups. If some legislators are more likely to toe the party group line than others, the

assumption that errors are identically and independently distributed is likely to be violated

(Rosenthal and Voeten 2004). More importantly, as will be shown below, a two-dimensional

spatial model fits the observed vote choices almost perfectly. When error rates are so low,

parametric methods push ideal points to the edge of the policy space to maximize the log-

likelihood (Armstrong et al. 2014: 266; Rosenthal and Voeten 2004). As a result, distances

between ideal points do not only reflect differences in policy positions, but also differences in

the number of voting errors. This ‘sag’ problem exaggerates distances of legislators with

extreme and interior policy positions (Armstrong et al. 2014: 269). One cost of a non-

parametric method like OC is that the ideal point scaling results do not come with estimates

of the associated uncertainty, but its benefits in terms of more valid ideal point estimates

outweigh this shortcoming.

To investigate possible changes in the number and content of conflict dimensions over

time, I apply the OC algorithm separately to each legislative term. Table 2 provides an

overview of the input data and various model fit statistics. The analysis is limited to

legislators who voted in at least 15 roll calls in a legislative term and excludes extremely lop-

sided votes where the minority consisted of less than 2.5 per cent of voters. Unfortunately,

these restrictions are necessary to yield meaningful estimates. As the top part of Table 2

shows, the consequences of the application of these criteria are more pronounced for

13

legislators than for roll calls. The roll call vote criterion results in a ‘loss’ of only a single

vote in the sixth legislative term. Yet, a considerable number of legislators are dropped in

each term because they have not taken part in the minimal number of roll call votes.9

Dimensionality of the conflict space

The first step of the analysis examines the dimensionality of the political space separately for

each legislative term. Unfortunately, no single best method exists to determine the ‘correct’

number of conflict dimensions. To make sure the selection of dimensions is robust and

provides substantively meaningful results, several statistical criteria were combined. First, as

recommended by Poole (2005: 144), scree plots of the eigenvalues of the double-centred

agreement score matrix were examined. Table 2 provides the number of eigenvalues after

which the plot flattens out.10

None of the scree plots indicates a dimensionality larger than

three. Thus, in a second step, one, two, and three-dimensional scaling solutions were

generated with the OC algorithm. The fit statistics for these solutions in the lower part of

Table 2 provide further information about the number of dimensions underlying legislative

voting in the Reichstag. To judge the model fit of OC, we can look at the percentage of vote

choices that were correctly classified (PCC) and the aggregate proportional reduction in error

(APRE) across all roll calls. The latter measure has the advantage that it takes into account

the distribution of vote margins (Poole 2005: 129). APRE provides an indication of the extent

to which the OC solution reduces the classification error vis-à-vis a theory-free baseline

model that assumes that all legislators vote with the majority.

9 A replication using a threshold of 10 roll call votes does not yield meaningful results. The results regarding

dimensionality and the relative location of party groups in the political space are similar when a threshold of 20

or 25 roll call votes is used. 10

See Figure A1 in the online appendix for the actual plots.

14

Table 2 Fit of one- to three-dimensional Optimal Classification models by

legislative term

NDRT RT 1 RT 2 RT 4 RT 5 RT 6 RT 7

Total no. of roll calls 81 48 42 41 47 33 30

No. of scaled roll calls 81 48 42 41 47 32 30

Total no. of legislators 338 411 421 444 426 415 429

No. of scaled legislators 290 356 370 357 371 352 312

No. of eigenvalues (scree plots) 1 2 1 2 2 3 3

One dimension

Percentage correctly classified 94.9 87.8 95.6 89.9 93.3 93.4 92.1

Aggregate PRE 87.4 68.1 88.1 73.9 83.3 84 76.7

Two dimensions

Percentage correctly classified 96.6 93.5 97.6 96.7 98.7 98.5 99

Aggregate PRE 91.7 83.1 93.6 91.6 96.8 96.4 96.9

Difference correct classification 1.7 5.7 2 6.8 5.4 5.1 6.9

Difference aggregate PRE 4.3 15 5.5 17.7 13.5 12.4 20.2

No. roll calls PRE diff. > 0.20 4 16 3 8 11 6 11

% roll calls PRE diff. > 0.20 4.9 33.3 7.1 19.5 23.4 18.2 36.7

Three dimensions

Percentage correctly classified 97.6 95.8 98.2 98.3 99.1 99 99.5

Aggregate PRE 94.1 88.9 95 95.6 97.9 97.6 98.4

Difference correct classification 1 2.3 0.6 1.6 0.4 0.5 0.5

Difference aggregate PRE 2.4 5.8 1.4 4 1.1 1.2 1.5

No. roll calls PRE diff. > 0.20 2 5 0 2 0 0 0

% roll calls PRE diff. > 0.20 2.5 10.4 0 4.9 0 0 0

Conclusion (no. of dimensions) 1+ 2+ 1+ 2+ 2 2 2

Notes: NDRT=Reichstag of North German Confederation, RT=Reichstag of German Empire, PRE=Proportional

Reduction in Error

The first thing to notice about the model fit statistics is that the correct classification rates for

all legislative terms are very high. For one-dimensional models, they range between 87.8 per

cent in RT 1 and 95.6 in RT 2. Yet for most terms, adding a second dimension still improves

the classification rate considerably. For the two-dimensional solutions, the rates vary from

93.5 per cent in RT 1 to 99 per cent in RT 7. Classification rates as high as these leave little

room for further model fit improvements as a result of the introduction of a third dimension.

In fact, we can be quite certain that political conflict in the NDRT, RT 2, as well as RT 5 to 7

can be represented in fewer than three dimensions. For these terms, the difference in the rate

of correct classification between a three- and two-dimensional solution is one percentage

point or lower, and the difference in APRE is less than 5 percentage points. Except for a

couple of roll calls in the NDRT, none of the proportional reduction in error (PRE) values for

15

individual roll calls improves by more than 20 per cent when introducing a third dimension.

The model fit improvements for RT 1 and 4 are somewhat larger but still rather marginal.

Using a 5-percentage point increase in the rate of correct classification, a 10-percentage

point increase in APRE, and a PRE increase of more than 20 per cent for at least 10 per cent

of the roll calls as a combined threshold for judging the gain in predictive power of adding a

second dimension, we can conclude that political conflict in RT 1 and RT 4 to 7 was at least

two-dimensional. The NDRT and RT 2 show some gain in predictive power when estimating

a second dimension as well, but compared to the other terms, it is rather negligible. In

summary, the results suggest that political conflict in RT 5, 6, and 7 was almost certainly

two-dimensional. Taking into account the number of dimensions indicated by the scree plots,

the balance of evidence suggests that political conflict in RT 1 and 4 was also two-

dimensional, but political conflict in the NDRT and RT 2 was one-dimensional.

Nature of conflict

The content of conflict dimensions is determined by substantive knowledge rather than

statistical methods. Such knowledge about the content of votes also helps to further interpret

the dimensionality of the conflict space. Figure 3 plots the OC solutions for all legislative

terms. Depending on the number of dimensions selected based on the statistical criteria

outlined in the previous section, the panels show either the estimated legislator ideal points of

the one- (NDRT and RT 2) or two-dimensional OC solution (all other terms). The

interpretation of the content of the political conflict space of some of the legislative terms is

somewhat complicated by the fact that the mathematical dimensions do not align with the

substantive dimensions (for a similar result in the case of the National Assembly of the

French Fourth Republic, see Poole 2005: 152; Rosenthal and Voeten 2004). For ease of

comparison and interpretation, the two-dimensional solutions were rotated to ensure that the

centroid of the German Conservative legislators’ ideal points aligned with the diagonal in the

top-right quadrant of the plot. Little doubt exists that the German Conservatives held

conservative positions on both socio-economic and regime issues. To learn about the content

of issue dimensions, we can examine by how much the PRE value for a roll call increases as

a result of the estimation of an additional dimension.

16

Figure 3 Legislator ideal point estimates by legislative terms

Notes: NDRT=Reichstag of North German Confederation, RT=Reichstag of German Empire, 1D=one-

dimensional solution, 2D=two-dimensional solution, rotated to fix centroid of German Conservatives to

diagonal line in upper right quadrant. Z=Centre Party, K=German Conservatives, R=Free Conservatives,

N=National Liberals, F=Left Liberals, S=Socialists, M=Minorities, W=Independents, B=Particularists (NDRT

only), A=Liberal Centre (NDRT only), L=Liberal Empire Party (RT 1 only), V=Liberal Association (RT 5

only).

17

In the NDRT, the few roll calls whose PRE value increases substantively after allowing for

the estimation of a second dimension are quite diverse and cover similar topics to those

whose PRE values do not increase substantively. Thus, in isolation, the approach does not

provide insights for the interpretation of the substance of the one-dimensional conflict space

of this term. In the case of RT 1, comparing roll calls with a small increase in PRE values (<

5 per cent) with roll calls with a large increase (> 20 per cent) indicates that the first

dimension captures differences in views about civil rights and liberties as well as the powers

of parliament, especially its control over the military budget, whereas the second dimension

captures differences in views about the relation between the state and the church, the

distribution of competences of regional states vice-a-vis the federal government, and the

setting of tariffs. In short, the first dimension bundles ‘regime’ matters, while the second

dimension covers ‘economic and social policy’. With respect to the latter, the term ‘social’

refers to ideas about the (federal) state’s role in society, rather than exclusively redistributive

policies.

As the panel for RT 1 in Figure 3 shows, the regime dimension pits left liberals (F), the

Centre Party (Z), and various regional and ethnic minorities (M), who champion civil rights

and liberties and demand increased parliamentary powers and control over the military,

against the conservative camp (K, F), which defends the current regime. National Liberals

(N) and members of the Liberal Empire Party (L, which existed as a party group only in

RT 1) take somewhat intermediate positions. In contrast, the economic and social policy

dimension differentiates the conservative parties, the Centre Party, and the minorities on the

one side from the three liberal party groups on the other side. The former defend traditional

economic structures, protectionist trade policies, existing social hierarchies, as well as the

independence of the Reich’s component states, while the latter are proponents of deregulated

markets, free trade, the abolishment of established feudal prerogatives, and a strong nation

state.

This pattern of positions is relatively consistent throughout all legislative terms with

two-dimensional solutions presented in Figure 3 (i.e. RT 1 and RT 4 to 7). On one dimension,

the positions of members of the conservative groups and the Centre Party are aligned with

each other against the positions of members of the liberal and socialist groups; on the other

dimension, the positions of members of the conservative groups and the National Liberals are

aligned with each other against the positions of members of the Centre Party, Socialists, and

Left Liberal groups. Of course, the locations of ideal points and the distances between them

are not comparable across legislative terms, and, although reasonable, the rotation of the plots

18

is somewhat arbitrary. Yet, all two-dimensional solutions show the same broad constellation

of party group members’ ideal points. Starting with the German Conservatives in the upper

right quadrant and moving in clock-wise direction, the Free Conservatives are the closest,

followed by the National Liberals, the Left Liberals, and Socialists. Finally, the minority

groups and the Centre Party members complete the circle.

In terms of face validity, the scaling results are in line with a number of stylized

historical facts about coalition and voting patterns in the Bismarck era: The Free

Conservatives formed a bridge between National Liberals and German Conservatives, the

minority groups tended to align with the Centre Party, and the left wing of the National

Liberals, which seceded in 1880 and formed the Liberal Association in RT 5, held

intermediate positions between the left liberal Progressive Party and the remaining members

of the National Liberal Party. Furthermore, Figure A2 in the online appendix shows that this

division within the National Liberal Party is visible in earlier legislative terms as well. Left-

wing National Liberals, who would later secede from the party, tended to hold positions

closer to those of left liberals than the bulk of their fellow group members.

Even the apparently one-dimensional conflict space of RT 2 can be interpreted in that

vein. In RT 2, the first dimension plotted in Figure 3 represents different views on regime

matters, with the Centre Party, the minority groups, the Socialists, and the Left Liberals

favouring democratic reforms, while the National Liberals, the Free Conservatives and the

German Conservatives are protecting the status quo. The second, socio-economic policy

dimension does not add much to the statistical model fit (and thus only the one-dimensional

solution is shown in Figure 3). Still, some indications exist that the second dimension is

based on the same type of divisions as in the clearly two-dimensional legislative terms. The

few roll calls with a considerably improved PRE value relate to tariffs and the building of a

new plenary building for the Reichstag (i.e. trade policy and powers of the federal state).

However, most other roll calls in this term were about matters relating to the regulation of the

press, the military, the organisation of Courts, and the proceedings in criminal trials. In other

words, with matters relating to the regime dimension. Thus, the apparent reversal to a single

dimension in RT 2 is an artefact of a legislative agenda dominated by regime matters rather

than real changes in the underlying conflict space. Disagreements over socio-economic issues

did not disappear in RT 2, they were just hardly visible in parliamentary activities.

Unlike in RT 1 and 2, the statistically dominant dimension of conflict in RT 4 was the

socio-economic dimension, and the regime dimension became secondary. This signifies

another but less consequential agenda effect: the change in emphasis in law-making from

19

state-building to policy-making after 1878. In RT 4, the socio-economic dimension captures

disagreements about the tariff law, workers’ accidence insurance, trade and industry

regulations, and taxes. The regime dimension reflects mainly divisions about the adoption of

the anti-socialist law and the military law. The regime dimension in RT 5 represents

disagreements about matters like the renewal of the anti-socialist law, the abolition of the

clergy expatriation law, and the official language of the state assembly of Alsace-Lorraine.

The socio-economic dimension is dominated by disagreements about the tariff law, the

domestic commerce regulation, the workers’ health insurance law, and the accidence

insurance law.

In the political conflict space of RT 6, the socio-economic dimension is again

dominated by disagreements about the tariff law. It also captures different views about an

amendment of an international trade treaty with Spain, as well as the establishment of a state

spirits monopoly. Next to divisions on the clergy expatriation law, the extension of the anti-

socialist law, and the size of the military, the regime dimension also reflects disagreement

about the country’s new colonial ambitions. Finally, disagreements about tariffs still factored

into the socio-economic dimension of RT 7, but the dimension also reflects divergent views

on domestic commerce and bank regulation. The regime dimension is formed by divisions

about the size of the military, the length of the parliamentary term, the renewed extension of

the anti-socialist law, and the publicity of Court proceedings. Statistically, the regime

dimension was the dominant one again in RT 7, which might reflect the partly reactionary

agenda of the Kartell coalition that dominated the Reichstag during that term. Somewhat

inconsistently, the regime dimension also reflects disagreement about the old age and

disability insurance bill. All previous votes on Bismarck’s social insurance bills were related

to the socio-economic dimension. However, this is the only instance of a roll call vote clearly

not conforming to the general pattern.

After having seen how different subject matters relate to distinct dimensions in other

legislative terms, we can also revisit the content of the dominant dimension in the NDRT. In

contrast to the one-dimensional solution found for RT 2, which results mainly from the

selective agenda voted upon during that term, the single dominant dimension of the NDRT

included votes on both regime and socio-economic issues. Thus, legislators’ positions on

these two issue bundles were tightly aligned in the NDRT, but diverged throughout all later

legislative terms of the Reichstag.

20

Discussion

This study examined political conflict dimensions in Imperial Germany during Chancellor

Bismarck’s term of office. Based on an analysis of roll call votes taken in the Reichstag

between 1867 and 1890, the dimensional scaling analysis suggests that throughout the period

considered, two bundles of issues can be distinguished. The first bundle relates to socio-

economic issues, including the role of the federal state and the Church in governing society,

the regulation of economic activities, tax and redistributive policies. The second bundle refers

to regime matters, including the relative power and resources of parliament, civil rights and

liberties, electoral provisions, and the organisation of justice. In the North German

Confederation, political views of legislators along both issue bundles where largely aligned

and thus collapsed into a single liberal-conservative dimension. However, the appearance of

the Catholic Centre Party after the foundation of the German Empire in 1871 reshaped the

political conflict space. In terms of social and economic issues, the Centre Party supported

conservative solutions. Yet, in light of the minority status of Catholics in Germany and the

repressive policies pursued against the Catholic Church during the so-called ‘Culture War’ in

the early 1870s, it defended liberal positions with respect to civil rights and liberties as well

as the powers of parliament. The emergence of the Centre Party, together with a move of

parts of the National Liberals towards more conservative positions on regime issues, created a

two-dimensional space that characterized political conflict throughout the remainder of the

study period. Interestingly, these two issue bundles, socio-economic policies and regime

matters, continued to play a major role in structuring political conflict in Germany, not only

during the Wilhelmine era of the Kaiserreich (Smith and Turner 1981), but also in the

Weimar Republic (Debus and Hansen 2010; Hansen and Debus 2012).

Methodologically, the study illustrates the importance of taking changes in the agenda

into account when using roll call votes to assess the dimensionality of a political conflict

space. Based on purely statistical criteria for the selection of dimensions, we would have

concluded that opinions on socio-economic issues did not divide legislators in the second

legislative term of the Reichstag from 1874 to 1877. However, a closer look at the substance

of legislation voted upon during that term revealed that this finding was a result of the very

small number of votes taken on these types of issues. Opposing positions on socio-economic

issues had not suddenly converged; they just had little opportunity to manifest themselves in

parliamentary activities. In the second Reichstag term, these activities focused on issues

related to the regime dimension. Especially in historical cases with comparatively modest

21

legislative productivity, changes in the legislative agenda can have undue effects on the

results of a roll call analyses; their interpretation needs to be supported by information about

the policy content of votes.

The Centre Party’s electoral success in 1871 was based on the activation of a religious

cleavage dividing the Protestant majority and the Catholic minority in Germany. At least in

Prussia, Catholics were generally economically disadvantaged compared to Protestants.

However, economic inequality did not play a primary role in the mobilisation of Catholic

voters. Centre voters came from all economic strata (Nipperdey 2013: 344). Partly as a result

of this economic diversity in the support base, especially the Centre Party’s early electoral

announcements were vague on socio-economic issues, highlighting the defence of Church

prerogatives and the federal nature of the Reich (Lepper 1998). The latter implied

maintaining limits on the power of Protestant-dominated Prussia. In short, the Centre Party’s

continuing electoral success was based on a genuinely religious divide that cut across

economic sectors and social strata. However, its sudden appearance in 1870/71 also

demonstrates that the political entrepreneurship of Catholic politicians and indeed the

mobilization of Catholic voters by the clergy were instrumental for activating this cleavage

(Anderson 1988: 134–136; Anderson 2000: 69–105). Once the Centre Party had been

established as an important electoral and parliamentary force, it played a major role in

perpetuating this division.

Thus, in the case of Imperial Germany, an exclusive focus on the role of economic

inequality as a factor influencing regime stability and change (Acemoglu and Robinson 2006;

Ansell and Samuels 2014; Boix and Svolik 2013) misses an important part of the country’s

political history. Before the emergence of the Centre Party, the actor and interest

constellation in the German party system resembled quite closely the elite-competition

interaction theorised by Ansell and Samuels (2014). An economically rising group,

represented by the liberals, demanded more democratic participation and rights, possibly to

protect their newly gained wealth from expropriation by the incumbent elite, represented by

the government and the conservatives. However, the emergence of the Centre Party at the

beginning of the 1870s does not fit this narrative.

The Centre Party defended existing democratic achievements and promoted progressive

reforms to defend the prerogatives of the Catholic Church and the Catholic ‘way of life’

against the Protestant-dominated governments of the Reich and Prussia. The fact that the

Centre Party regularly supported the same socio-economic policies as the conservatives

provides clear evidence that its regime preferences, which ran counter to those of

22

conservatives, cannot be reduced to economic considerations; and given its electoral strength

and often pivotal position in the Reichstag, the Centre Party played a major role in

determining the political development of Germany. In fact, when Bismarck decided to

reverse course in the second half of the 1870s to pursue more protectionist economic policies

that favoured agricultural landholders and large industrialists, the Centre Party provided the

required support in the Reichstag when the National Liberals refused. Indeed, Bismarck’s

ability to govern with varying majorities because of the Centre Party’s parliamentary strength

and position in the political space may have bolstered the stability of the regime by avoiding

legislative gridlock and a direct bipolar confrontation between the liberal and conservative

camp. However, a more thorough investigation of this possibility is a topic for future

research. As a more immediate conclusion, the study points to the need to delineate clearly

the temporal and geographical scope conditions of theories of regime stability and change.

When theoretical arguments presume a particular actor and interest constellation, it is

difficult to see how they can provide adequate explanations for cases where these

constellations were only partially present, only present during a particular period of time, or

indeed not present at all.

References

Acemoglu, D. and Robinson, J. A. (2006) Economic Origins of Dictatorship and Democracy,

Cambridge: Cambridge University Press.

Alemán, E. and Saiegh, S. (2014) ‘Political realignment and democratic breakdown in

Argentina, 1916-1930’, Party Politics 20(6): 849–863.

Anderson, M. L. (1988) Windthorst: Zentrumspolitiker und Gegenspieler Bismarcks,

Düsseldorf: Droste.

Anderson, M. L. (2000) Practicing Democracy: Elections and Political Culture in Imperial

Germany, Princeton: Princeton University Press.

Ansell, B. W. and Samuels, D. J. (2014) Inequality and Democratization: An Elite-

Competition Approach, Cambridge: Cambridge University Press.

Armstrong, D. A. et al. (2014) Analyzing Spatial Models of Choice and Judgment with R,

Boca Raton: CRC Press.

Barkin, K. D. (1987) ‘1878-1879 : The Second Founding of the Reich, a Perspective’,

German Studies Review 10(2): 219–235.

Biefang, A. (2012) Die andere Seite der Macht: Reichstag und Öffentlichkeit im ‘System

Bismarck’, Düsseldorf: Droste.

23

Boix, C. (2003) Democracy and Redistribution, Cambridge: Cambridge University Press.

Boix, C. and Svolik, M. W. (2013) ‘The Foundations of Limited Authoritarian Government:

Institutions, Commitment, and Power-Sharing in Dictatorships’, Journal of Politics

75(2): 300–316.

Butzer, H. (1999) Diäten und Freifahrt im Deutschen Reichstag: Der Weg zum

Entschädigungsgesetz von 1906 und die Nachwirkung dieser Regelung bis in die Zeit

des Grundgesetzes, Düsseldorf: Droste.

Carrubba, C. J. et al. (2006) ‘Off the Record: Unrecorded Legislative Votes, Selection Bias

and Roll-Call Vote Analysis’, British Journal of Political Science 36: 691.

Clinton, J., Jackman, S. and Rivers, D. (2004) ‘The Statistical Analysis of Roll Call Data’,

American Political Science Review 98(2): 355–370.

Debus, M. and Hansen, M. E. (2010) ‘Die Dimensionalität der Reichstage der Weimarer

Republik von 1920 bis 1932’, Politische Vierteljahresschrift 51: 15–42.

Gerschenkron, A. (1943) Bread and Democracy in Germany, Ithaca: Cornell University

Press.

Hansen, M. E. and Debus, M. (2012) ‘The behaviour of political parties and MPs in the

parliaments of the Weimar Republic’, Party Politics 18(5): 709–726.

Lepper, H. (1998) Volk, Kirche und Vaterland: Wahlaufrufe, Aufrufe, Satzungen und Statuten

des Zentrums, 1870-1933, Düsseldorf: Droste.

Lipset SM, Rokkan S (1967) ‘Cleavage structures, party systems and voter alignments: an

introduction’, in SM Lipset and S Rokkan (eds.) Party Systems and Voter Alignments:

Cross-National Perspectives. New York: Free Press.

Moore, B. (1966) Social Origins of Dictatorship and Democracy: Lord and Peasant in the

Making of the Modern World, Harmondsworth: Penguin University Books.

Michels R (1911) Zur Soziologie des Parteiwesens in der modernen Demokratie:

Untersuchungen über die oligarchischen Tendenzen des Gruppenlebens. Leipzig:

Verlag von Dr. Werner Klinkhardt.

Nipperdey, T. (2013) Deutsche Geschichte 1866-1918: Band II, Machstaat vor der

Demokratie, München: C.H. Beck.

Pflanze, O. (1982) ‘Bismarck’s Herrschaftstechnik als problem der gegenwärtigen

Hisoriographie’, Historische Zeitschrift 234(3): 561–599.

Pollmann, K. E. (1985) Parlamentarismus im Norddeutschen Bund 1867-1870, Düsseldorf:

Droste.

24

Poole, K. T. (2000) ‘Nonparametric Unfolding of Binary Choice Data’, Political Analysis

8(2): 211–237.

Poole, K. T. (2005) Spatial Models of Parliamentary Voting, Cambridge: Cambridge

University Press.

Poole, K. T. and Rosenthal, H. L. (2007) Ideology and Congress, New Brunswick:

Transaction Publishers.

Rauh, M. (1973) Föderalismus und Parlamentarismus im Wilhelminischen Reich,

Düsseldorf: Droste.

Reichstagsbureau (1896) Generalregister zu den stenographischen Berichten über die

Verhandlungen und den amtlichen Drucksachen des konstituirenden Reichstages, des

Reichstages des Norddeutschen Bundes, des Deutschen Zollparlaments und des

Deutschen Reichstages, Berlin: Norddeutsche Buchdruckerei und Verlags-Anstalt.

Rosenthal, H. and Voeten, E. (2004) ‘Analyzing Roll Calls with Perfect Spatial Voting:

France 1946-1958’, American Journal of Political Science 48(3): 620–632.

Schauff, J. (1973) ‘Das Wahlsystem des Deutschen Reiches und die Zentrumspartei’, in G. A.

Ritter (ed.). Die Deutschen Parteien vor 1918. Köln: Kiepenheuer & Witsch, pp. 299–

309.

Smith, W. and Turner, S. (1981) ‘Legislative Behavior in the German Reichstag, 1898–

1906’, Central European History 14(1): 3–29.

Stalmann, V. (2000) Die Partei Bismarcks: Die Deutsche Reichs- und Freikonservative

Partei, 1866-1890, Düsseldorf: Droste.

1

Political conflict in Bismarck’s Germany: An analysis of

parliamentary voting, 1867-1890

Supplementary information

Appendix 1 Additional tables and figure

• Table A1: Share of Reichstag seats of party groups by legislative term

• Table A2: Policy area of roll call vote by legislative term

• Table A3: Type of subject matter of roll call vote by legislative term

• Figure A1: Eigenvalue scree plots by legislative term

Appendix 2 Detailed description of data sources, collection, and coding

Appendix 3 Face validity check: Intra-group divisions in the National Liberal Party

2

Appendix 1 Additional tables and figure

Table A1 Share of Reichstag seats of party groups by legislative term

NDRT

1867-

1871

RT 1

1871-

1874

RT 2

1874-

1877

RT 3

1877-

1878

RT 4

1878-

1881

RT 5

1881-

1884

RT 6

1884-

1887

RT 7

1887-

1890

Total

(German) Conservatives 69 54 21 40 59 50 77 79 449

0.23 0.14 0.05 0.10 0.15 0.13 0.19 0.20 0.15

Free Conservatives 34 38 32 38 56 27 28 41 294

0.11 0.10 0.08 0.10 0.14 0.07 0.07 0.10 0.10

French minority 15 15 15 15 15 15 90

0.04 0.04 0.04 0.04 0.04 0.04 0.03

Guelphs 7 4 5 10 10 9 4 49

0.02 0.01 0.01 0.03 0.03 0.02 0.01 0.02

Polish minority 11 13 14 14 14 18 16 13 113

0.04 0.03 0.04 0.04 0.04 0.05 0.04 0.03 0.04

Centre Party 60 91 92 94 99 101 98 635

0.16 0.23 0.23 0.24 0.25 0.25 0.25 0.21

Federalist-Constitutionalist 18 18

Association 0.06 0.01

Liberal Centre Party 13 13

0.04 0.00

Liberal Empire Party 30 30

0.08 0.01

National Liberals 83 120 151 128 99 47 50 99 777

0.28 0.31 0.38 0.32 0.25 0.12 0.13 0.25 0.25

Independents 18 13 10 9 12 6 5 5 78

0.06 0.03 0.03 0.02 0.03 0.02 0.01 0.01 0.03

Free Association 15 15

0.05 0.00

Liberal Association 46 46

0.12 0.02

Progress Party / 30 44 49 41 26 59 65 32 346

German Liberal Party 0.10 0.12 0.12 0.10 0.07 0.15 0.16 0.08 0.11

German People's Party 1 1 3 3 8 7 0 23

0.00 0.00 0.01 0.01 0.02 0.02 0.00 0.01

Socialist Workers’ Party 6 2 9 12 9 12 24 11 85

0.02 0.01 0.02 0.03 0.02 0.03 0.06 0.03 0.03

Total members 297 382 397 397 397 397 397 397 3061

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

Notes: Cells are shaded according to relative frequency. NDRT=Reichstag of North German Confederation,

RT=Reichstag of German Empire. Socialist Workers’ Party refers to seats of the General German Workers’

Association and the Social Democratic Workers’ Party before RT 3.

3

Table A2 Policy area of roll call vote by legislative term

Policy area NDRT RT 1 RT 2 RT 3 RT 4 RT 5 RT 6 RT 7 Total

Social Welfare 0 0 0 0 3 8 0 4 15

0.00 0.00 0.00 0.00 0.07 0.17 0.00 0.13 0.05

Domestic Macroeconomic Issues 6 0 0 0 4 4 4 2 20

0.07 0.00 0.00 0.00 0.10 0.09 0.12 0.07 0.06

Foreign Trade 1 2 1 1 18 4 21 5 53

0.01 0.04 0.02 0.11 0.44 0.09 0.64 0.17 0.16

Banking, Finance, and Domestic 8 2 0 1 4 15 0 8 38

Commerce 0.10 0.04 0.00 0.11 0.10 0.32 0.00 0.27 0.11

Civil Rights, Minority Issues, and 5 9 1 2 4 7 2 4 34

Civil Liberties 0.06 0.19 0.02 0.22 0.10 0.15 0.06 0.13 0.10

Defence 8 12 5 0 4 7 2 4 42

0.10 0.25 0.12 0.00 0.10 0.15 0.06 0.13 0.13

Law, Crime, and Family Issues 25 1 12 1 1 0 0 1 41

0.31 0.02 0.29 0.11 0.02 0.00 0.00 0.03 0.12

Government Operations 20 20 12 4 2 2 1 1 62

0.25 0.42 0.29 0.44 0.05 0.04 0.03 0.03 0.19

Other Issues 8 2 11 0 1 0 3 1 26

0.10 0.04 0.26 0.00 0.02 0.00 0.09 0.03 0.08

Total 81 48 42 9 41 47 33 30 331

0.24 0.15 0.13 0.03 0.12 0.14 0.10 0.09

Notes: Cells are shaded according to relative frequency. NDRT=Reichstag of North German Confederation,

RT=Reichstag of German Empire. Policies classified according to the classification scheme of the Comparative

Agendas Project (http://www.comparativeagendas.info).

4

Table A3 Type of subject matter of roll call vote by legislative term

Type NDRT RT 1 RT 2 RT 3 RT 4 RT 5 RT 6 RT 7 Total

Legislation 54 30 33 9 35 38 28 29 256

0.67 0.62 0.79 1.00 0.85 0.81 0.85 0.97 0.77

Constitutional 5 8 5 0 1 0 1 1 21

amendment 0.06 0.17 0.12 0.00 0.02 0.00 0.03 0.03 0.06

Budget 5 1 1 0 3 7 3 0 20

0.06 0.02 0.02 0.00 0.07 0.15 0.09 0.00 0.06

Request to 8 5 1 0 0 0 0 0 14

government 0.10 0.10 0.02 0.00 0.00 0.00 0.00 0.00 0.04

Validity of 3 2 0 0 1 1 0 0 7

mandate 0.04 0.04 0.00 0.00 0.02 0.02 0.00 0.00 0.02

Parliamentary 2 1 2 0 0 1 0 0 6

resolution 0.02 0.02 0.05 0.00 0.00 0.02 0.00 0.00 0.02

International 2 0 0 0 1 0 1 0 4

treaty 0.02 0.00 0.00 0.00 0.02 0.00 0.03 0.00 0.01

Address to 2 1 0 0 0 0 0 0 3

Kaiser 0.02 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.01

Total 81 48 42 9 41 47 33 30 331

0.24 0.15 0.13 0.03 0.12 0.14 0.10 0.09

Of which:

Procedural 2 5 2 0 1 5 5 2 22

0.02 0.10 0.05 0.00 0.02 0.11 0.15 0.07 0.07

Final passage 2 2 5 2 6 2 4 7 30

0.02 0.04 0.12 0.22 0.15 0.04 0.12 0.23 0.09

Notes: Cells are shaded according to relative frequency. NDRT=Reichstag of North German Confederation,

RT=Reichstag of German Empire.

Table A3 shows that the roll call data are dominated by votes on substantively important and

legally consequential matters. Overall, almost 8 in 10 roll call votes (77 per cent) took place

in the context of a legislative procedure, and another 6 per cent each referred to constitutional

amendments and budget matters. The combined proportion of these three types of subject

matters was somewhat lower during the NDRT and RT 1. The possibility that this difference

is due to the lack of an alternative procedure to count votes when they were too close to call

before the introduction of voting by division in 1874 cannot be ruled out. Still, even in these

two terms, votes on the three types of subject matters dominated the roll call data (79 per cent

in the NDRT, and 81 per cent in RT 1). In general, the overwhelming majority of roll call

votes (93 per cent) has been called on substantive, not procedural matters. Finally, most roll

call votes (91 per cent) are on specific parliamentary decisions or bill amendments rather than

final passage votes of legislation. This is not a problem for scaling policy positions of

legislators, as votes on amendments are more informative than final passage votes if the latter

are based on logrolls or package deals.

5

Figure A1 Eigenvalue scree plots by legislative terms

Notes: Eigenvalues of the double-centred agreement score matrix. The number of dimensions before the ‘elbow’

in the number of eigenvalues indicates the dimensionality of the data. NDRT=Reichstag of North German

Confederation, RT=Reichstag of German Empire.

6

Appendix 2 Detailed description of data sources, collection and coding

The source of the voting information is the overview of roll call votes in Appendix A of the

‘General Register for the Stenographical Reports of all Reichstag Sessions from 1867 to

1895’ (Reichstagsbureau 1896). After downloading the digitized version of the General

Register from the website of the Bavarian State Library, optical character recognition (OCR)

software was used to convert the PDF images of the roll call vote appendix into machine-

readable text.12

To identify the full population of legislators and disambiguate their party

group affiliation and electoral district at the time of a particular vote, the vote data was linked

to biographical information from the online database Biorab-Kaiserreich, which is hosted by

the Centre for Historical Social Research at the GESIS Leibnitz Institute for the Social

Sciences.13

Before combining the two datasets through a custom-made record linkage

algorithm, information from both sources was extracted through computer scripts developed

in Python. Several steps in the data collection process required extensive manual review and

corrections based on the consultation of historical primary and secondary sources.

The roll call vote appendix of the General Register consists of two parts, which were

processed separately. The first part provides a chronologically numbered list of all roll call

votes, distinguishing votes by successive legislative terms and sessions. Each entry includes a

short description of the subject of the vote, when the vote was taken, and what the aggregate

outcome of the vote was. Where appropriate, it also includes a reference to the official

Reichstag document related to the vote. The second part provides an alphabetically numbered

list of Reichstag members, indicating, for each roll call vote they took part in, whether they

voted yeah, nay, or abstained. For each legislator, occupation, party group membership, and

electoral district are also recorded. Several party group affiliations are indicated if a legislator

changed his party group over time. The same is true for electoral districts. Unfortunately, no

information is provided about the timing of these changes.

Before further automated processing of the two lists, the results of the optical character

recognition (OCR) conversion were reviewed and spelling, punctuation, and formatting errors

12

For the download link, see http://www.reichstagsprotokolle.de/en_Band3_so_bsb00018728.html (last

accessed 9 April 2016). Because most free optical character recognition software does not have the capability to

recognize Fractur font, the commercial software ABBYY Recognition Server with Gothic/Fractur was used (see

http://www.frakturschrift.com/en:products:recognition_server [last accessed 9 April 2016). 13

See the database description at http://zhsf.gesis.org/ParlamentarierPortal/biorabkr.htm. Biographical

information about individual MdRs can be browsed at

http://zhsf.gesis.org/ParlamentarierPortal/biorabkr_db/biorabkr_db.php (both last accessed 11 April 2016)

7

were corrected.14

Python computer scripts were then used to extract the relevant information

from each of the lists and to reorganise it in the form of data matrices. From the list of roll

call votes, the scripts extracted information about the legislative term and session number in

which the vote took place, the number of yeah and nay votes, the name(s) of the sponsor(s) of

the bill or amendment, whether the vote was about a committee text, the date of the vote, the

reading in which the vote took place, the page numbers of the relevant session report, and the

number of associated session report appendix documents. Furthermore, the script identified

the type of issue under consideration, distinguishing between votes on laws (except

constitutional amendments and budget laws), constitutional amendments, budget matters,

requests to the Government (i.e. to the Bundesrat or Chancellor), the validity of mandates,

parliamentary resolutions, international treaty ratifications, and addresses to the Kaiser.

Again, the automated coding was manually checked and, if necessary, corrected.

The extraction of the voting information from the second part of the Appendix of the

General Register resulted in a dataset containing variables for the roll call number, vote

choice, first and last name of legislator, party group affiliation(s), electoral district(s), and

occupation. Once in machine-readable form, further consistency checks of the data were

conducted to ensure the OCR results were error-free. In particular, for each individual

Member of the Reichstag (MdR), it was confirmed that roll call numbers always appeared in

sequence and that they did not include duplicates. Another check made sure that vote choices

only included the possible options. Again, several corrections had to be made to the raw ORC

results to pass these consistency checks. The General Register and its appendices are

obviously based on the information in the official session reports. However, errors by the

Reichstagsbuero in collecting and processing that information cannot be ruled out. It is also

not clear whether the roll call vote appendix includes all MdRs, especially those that never

took part in a vote.

In order to identify the full population of MdRs over the 23 year period, to cross-

validate the attribution of votes to MdR names, and to differentiate multiple party group

affiliations and electoral districts over time, the data were merged with biographical

information about MdRs from the online database Biorab-Kaiserreich, which is hosted by the

Centre for Historical Social Research at the GESIS Leibnitz Institute for the Social

14

The OCR software provides a feature that is useful for efficiently reviewing the result of the conversion:

Elements of the converted text with a particularly low probability of having been recognized correctly are

highlighted. However, the feature is not error-free and cannot replace a careful reading of the converted text.

8

Sciences.15

To link the records in the voting data with the records in the bibliographical data,

a simple matching algorithm was developed. Both datasets include variables for the surname

of MDRs and their electoral districts. The algorithm loops through all legislative terms and

every member of the Reichstag during each term. Whereas the biographical data identifies a

singly electoral district for each MdR in a legislative term, the voting data only includes

variables indicating which electoral districts an MdR represented sometime during the study

period. Thus, in the first step, the algorithm attempts a fuzzy surname match of the MdR. If

the electoral district of the matched MdR from the biographical data matches one of the

electoral districts of the MdR in the voting data, the match is retained. If not, another fuzzy

surname match is attempted, but this time the biographical data is limited to only those MdRs

that were representing one of the electoral districts recorded for the MdR in the voting data.

In most instances, this algorithm resulted in an unambiguous and correct match. The

exceptions concern cases where several MdRs with the same surname were representing the

same district during the same legislative term. Another reason for incorrect merges were

compound names that occur as such in only one of the data sets. Since the fuzzy matching is

based on string distances of surnames, some long compound names have larger string

distances to one of their component names than to other, completely unrelated names. The

linkage algorithm does not take account of that possibility. These cases were manually

corrected after reviewing the merge results.

Observations that were missing in one or the other of the datasets - and thus could not

be merged - point to errors in one of them. Wherever such inconsistencies arose, third sources

were consulted to identify which dataset contained the correct information.16

By matching the

voting data with the biographical data, it was possible to add legislative term information for

those MdRs in the voting data that never voted, and to add legislative term information for

those that were members of several legislative terms, but only voted in some of them. MdRs

that did not accept their mandate or whose mandate ended before the first legislative session

of the term had started were dropped. Finally, it was possible to identify and correct ten errors

in the attribution of votes to MdRs in the General Register, mainly caused by a confusion of

identical surnames, and a couple of wrongly attributed electoral districts.

15

Again, computer scripts were developed in Python to automatically download the HTML pages of the

database, extract its information, and save it in a CSV file. See the database description at

http://zhsf.gesis.org/ParlamentarierPortal/biorabkr.htm. Biographical information about individual MdRs can be

browsed at http://zhsf.gesis.org/ParlamentarierPortal/biorabkr_db/biorabkr_db.php (both last accessed 11 April

2016). 16

Particularly useful were the reference books of Reichstag election outcomes by Phillips (1883) and Specht

and Schwabe (1904).

9

Like the electoral district information, membership in several party groups over time is

only indicated as such in the voting data without a specification of when a party group switch

occurred. Because the biographical data reports party group membership by legislative term,

it was also useful in coding party group membership of individual MdRs over time. If the

party group information in the two sources did not coincide, the party group information in

contemporary Reichstag handbooks as well as reference books on Reichstag elections

(Phillips 1883; Specht and Schwabe 1904) were consulted to identify the correct party group

affiliation.17

The party group information in the handbook was also used to manually review

the final party group coding of all MdRs.

Unfortunately, the biographical data does not provide the exact dates of Reichstag

membership for individual MdRs, but only the month of the year. In order to identify the

population of MdRs eligible to take part in a particular roll call vote, the precise dates of their

membership in the Reichstag is required. Starting with the rough periods provided by the

biographical data, membership start dates before the precise start date of the first legislative

session were replaced with the start date of the first legislative session. Similarly,

membership end dates after the precise end date of the last legislative session were replaced

with the end date of that session. For membership changes between those dates, Specht and

Schwabe (1904), Phillips (1883), and Reichstag session protocols and appendices were

consulted to identify the precise day of the membership change.

17

The 16 volumes of Georg Hirth’s semi-official ‘Parlaments Almanach’, published between 1867 and 1887,

can be accessed online (http://www.reichstagsprotokolle.de/en_rtbhzu.html, last accessed 12 April 2016). The

Reichstagsbuero started publishing an official handbook only in 1890.

10

Appendix 3 Face-validity check: Intra-group divisions in the National Liberal Party

As a further face validity check, Figure A2 replicates Figure 3 in the main text, but highlights

the positions of members of the left wing of the National Liberal Party that seceded in 1880,

formed the Liberal Association in 1881 (RT 5), and then formed a new party group with the

left liberals from 1884 (RT 6 and RT 7). The positions of these left-wing National Liberals

around Eduard Lasker, Ludwig Bamberger and Max von Forckenbeck are indicated by black

Xs. In line with historical accounts, members of the Liberal Association in RT 5 took

intermediate positions between the left liberal Progress Party and the National Liberals. With

the possible exception of RT 1, ideal points of future secessionists are generally closer to the

ideal points of left liberals than those of other National Liberals in earlier legislative terms as

well. This consistent finding provides significant support for the face validity of the scaling

results. Interestingly, once the secessionists had merged with the Progress Party, their voting

behaviour in RT 6 and RT 7 became even more distinct from the National Liberals but

indistinguishable to the voting behaviour of their new left liberal party group colleagues.

11

Figure A2 Legislator ideal point estimates by legislative terms

Notes: NDRT=Reichstag of North German Confederation, RT=Reichstag of German Empire, 1D=one-

dimensional solution, 2D=two-dimensional solution, rotated to fix centroid of German Conservatives to

diagonal line in upper right quadrant. Z=Centre Party, K=German Conservatives, R=Free Conservatives,

N=National Liberals, F=Left Liberals, S=Socialists, M=Minorities, W=Independents, B=Particularists (NDRT

only), A=Liberal Centre (NDRT only), L=Liberal Empire Party (RT 1 only), X=National Liberals, who seceded

from the party in 1880, formed the Liberal Association in RT 5, and formed a new party group with the left

liberals starting in RT 6.

12

References

Phillips, A. (1883) Die Reichstags-Wahlen von 1867 bis 1883: Statistik der Wahlen zum

Konstituierenden und Norddeutschen Reichstage, zum Zollparlament, sowie zu den fünf

ersten Legislatur-Perioden des Deutschen Reichstages, Berlin: Louis Gerschel.

Reichstagsbureau (1896) Generalregister zu den stenographischen Berichten über die

Verhandlungen und den amtlichen Drucksachen des konstituirenden Reichstages, des

Reichstages des Norddeutschen Bundes, des Deutschen Zollparlaments und des

Deutschen Reichstages, Berlin: Norddeutsche Buchdruckerei und Verlags-Anstalt.

Specht, F. and Schwabe, P. (1904) Die Reichstags-Wahlen von 1867 bis 1903: Eine Statistik

der Reichstagswahlen nebst den Programmen der Parteien un einem Verzeichnisse der

gewählten Abgeordneten, Berlin: Carl Heymanns.